FacimagePaperEx-150 10688life0904 974 982

Life Science J ournal 2012;9(4) http://www.lifesciencesite.
com

lifesciencej@gmail.com

974
Prediction of PEF and LITH logs using MRGC approach

Mahdi Pabakhsh
1
, Kamyar Ahmadi
2
*, Mohammad Ali Riahi
3
and Abbas Abbaszadeh Shahri
4

1
M.Sc. Student of Exploration Petroleum Engineering, Islamic Azad University, Science & Research Branch,
Tehran, Iran; m.pabakhsh.oil.eng@gmail.com
2
Academic staff and Faculty member, Department of Petroleum Engineering, Islamic Azad University, South
Tehran Branch, Tehran, Iran

3
Associate professor, Institute of Geophysics, University of Tehran, Tehran, Iran; mariachi@ut.ac.ir
4
Assistant professor, Department of Geophysics, Islamic Azad University, Hamedan branch, Hamedan, Iran;
a_abbaszadeh@iauh.ac.ir
*Corresponding Author: k.ahmadi@srbiau.ac.ir

Abstract: The Fuzzy logic method offers superior log estimation properties for a large class of well log functions
and has been employed as a standard tool in formation evaluation of Oil production zones. However it suffers from
spurious behavior in the vicinity of edge trends in log signals. In this article, we used The MULTI-RESOLUTION
GRAPH-BASED CLUSTERIN (MRGC) Supervised framework for obtaining lithology properties from PEF &
LITH logs that estimated in one of the Well (D) in the field since there is no core data in most wells. Estimations are
performed from basic information and model logs of another well (A), including RHOB, NPHI, DT, PHIE, and
NDS. Taking advantage of this framework, we show that it is feasible to recover log data from a relatively accurate
method especially in inhomogeneous formation than the Fuzzy logic method.
[Mahdi Pabakhsh, Kamyar Ahmadi, Mohammad Ali Riahi and Abbas Abbaszadeh Shahri. Prediction of PEF
and LITH logs using MRGC approach. Life Sci J 2012;9(4):974-982] (ISSN:1097-8135).
http://www.lifesciencesite.com. 150
Keywords: Pef log, Lith log, Estimate, MRGC, Fuzzy Logic

Introduction
In recent years, fuzzy logic and artificial neural
networks have been widely used for reservoir studies.
For example, several researchers (Saggaf & Nebrijs,
2003; Cuddy, 1997; Nordlund, 1996; Bois, 1984) have
applied neural networks and fuzzy logic in reservoir
studies of several fields.
Fuzzy logic was introduced in 1956 by Lotfi zadeh in a
paper entitled Fuzzy set. In the fuzzy logic, a
membership function is described which allows the
membership of more than one class with different
membership degrees.
Most previous researchers (Olson and Brill, 1997;
Bvbyk et al. 1999; Conte A. et al, 2006; Svtady Vieira,
2008) have attempted to determine PEF and LITH logs
but none of them have used MULTI-RESOLUTION
GRAPH-BASED CLUSTERIN (MRGC) method.
MRGC is one of the most important hierarchical
methods. In this research, MRGC approach is presented
to determine petrophysical parameters from well logs in
order to create different resolution of the data. In this
study, two logs from an oil field are used. In well A, the
PEF log is available and the LITH log is derived from
core data. These information are used to predict the
same logs in well D where PEF and LITH logs are not
available.
Methodology
MRGC method uses model logs and associated logs to
find a relationship between them to predict PEF and
LITH logs, so the estimated logs can be propagated to
the rest of the wells.
The MRGS method is, in fact, a combination of
artificial intelligence techniques and a hierarchical
clustering method. This method uses KRI and NI index
parameters which discern it from conventional methods.

WORKFLOW
Life Science J ournal 2012;9(4) http://www.lifesciencesite.com


975
Suppose there are 6 members, 3 clusters in a resolution
and 2 clusters in a higher resolution. In this case, the
MRGC method uses a parameter called KRI. In cross-
plots, especially in Neutron- Sonic where two points are
close to each other and there is no resemblance between
them, this method successfully distinguishes them. In
this approach, the following indices are added to make
the MRGC method more robust than other hierarchical
methods:
Neighborhood Index (Neighboring Index): This
parameter substitutes the distance parameter. As
mentioned before, when two points are close to each
other, they can be easily separated if they have high NI.
Unlike other hierarchical methods, depending on the
facies' behavior, the user can specify the number of
facies.
KRI Index: It is a combination of NI, distance and
weighted distance function M(x, y) which specifies the
Neighborhood or the degree of membership for M. If it
is low, it is affected by M; otherwise it has a high
membership degree and is not affected by M.
NI(X) = ) exp(
1
1
.

n
N
a n
m (1)
Where m the neighbor ranking, a is the resolution
parameter.

KRI=NI(x)M(x,y)D(x,y) (2)

In which M, is the weighting distance, D is the distance
between x and y.
First kernel or the center point which influences all of
its neighboring members" is specified", and then all the
members will be compared. The members that are
influenced by the kernel affect other members as well.
The boundaries are, therefore, specified where a
member is affected by its previous member but cannot
affect other members. So, the boundaries determine the
Split point and distinguish different groups based on the
parameters. According to this method, the model logs
(NDS, RHOB, NPHI, PHIE, and DT) are introduced
into the facimage section of the software to get the
clusters. The LITH log is then inputted as an associated
log and the data are trained.

Figure 1, at first, each log is divided into 2 clusters. Next, clustering is done based on the MRGC method with
minimum of 6 and maximum of 15 clusters.

Figure 1: Training the data's model logs as follow:
Model log:"RHOB, NDS, DT, PHIE, NPHI
Associated log: LITH log



976
Figure 2, 15 clusters have been selected to estimate the PEF log in Well A as a model, where the PEF log is
available. The estimated PEF log showed correlation with the actual PEF log in well A. figure 3, the PEF log is
propagated to the well D and the derived lithology is drawn.

Figure 4, the results of FUZZY method is shown in
which the RHOB, PHIE, NPHI, DT, and NDS logs are
entered and the PEF log is estimated. Clearly, the
predicted PEF log in well A with the Fuzzy method less
accurate than the one predicted with the MRGC method
where it shows a perfect correlation with the original
log.

Figure2.Estimated PEF LOG in A
well in red Color which has been
drawn 0 to 10 and it is having
correlation with PEF log in A well
in green color.
Figure 3.Estimated PEF log in D
well IN brown color which has
been drawn 0 to 10 and it is
having correlation with formation
of lithology.
Figure 4.Estimated PEF log by Fuzzy
method in black color which has been
drawn from 0 t0 10 compared with
estimated PEF log by MRGC method in
red color and available PEF log in A well.
In pointed circle, Fuzzy method has not
been operated perfectly and MRGC
method has been correlated completely
with available PEF log in A well.



977
Results and Discussion:
The Lithology of formation is heterogeneous and is
divided into dolomitic limestone, sandstone with
Anhydrate and shale which is accurately estimated from
LITH and PEF logs using MRGC method. The
Formation is divided into seven zones A1, A2, A3, A4,
A5, A6, and A7.
Shown in Figure 5 from left to right, the first track is the
derived lithology, the second track indicates LITH log
which is computed from core data and the last track is
the estimated LITH log from MRGC method which
shows a very good correlation.
Figure 3 shows well D where LITH and PEF logs are
not available. The obtained lithology from predicted
PEF log shows a good correlation with the obtained
lithology from petrophysical logs in Well D. As it can
be seen, the result of predicted PEF log is very reliable,
i.e., PEF number is 2 in the sandstone, 3 in shale, nearly
3 in limestone and 5 in anhydrite.
The estimated LITH log at well D in Figure 6 shows a
high correlation with PEF log and the derived lithology.

Figure 5. estimated LITH log by MRGC
method in left track compared with
LITH log obtained from Core in right
track in A well.
Figure 6.Estimated LITH and PEF log
by MRGC method in D well


978
Figure 7, well A and D are shown in a section which
demonstrates a good correlation between them and the
zone A6 is the sandstone reservoir zone.
Figure 8, Cross plot of the PEF log and the sonic logs in
well A is drawn, in which sonic log data are plotted in
the horizontal axis from 40 to 140 microseconds. The
colored data, green and yellow points, are the gamma
ray log from 0 to 200 API, representing clean sands,
shales and shaly sands respectively.
A linear relationship between DT and PEF logs is
established as PEF=9.70276-0.0906589DT and CC=
0.82 which is a high correlation between these logs. As
the transit time in the formation is decreased, the
corresponding PEF log value is increased.
In this case, the rocks are more condensate and porosity
is decreased. The PEF log value approaches 5 to 5.5
which is an indicative of limestone or anhydrate.
However, when the sonic log transit time is increased,
the PEF log reads a lower value and the formation is
more porous.

Figure 7. The section of A and B wells and the pointed correlation 7 Zones .

The range 2 to 3 of PEF log refers to sandstone to shale
and dolomite (Figure 8). The negative slope of the
regression line indicates an inverse relationship between
Sonic and PEF log.
Figure 9, the Cross-plot of PEF and neutron logs in
well A is shown. Neutron log data are plotted from -0.1
to 0.4 in the horizontal axis and the colored data refer to
gamma ray from 0 to 200 API in which the green and
yellow points are clean sands, shaly sands and shales
respectively. The linear relationship between NPHI and


979
PEF logs is PEF=5.76276-15.0999NPHI in which CC=
0.69 shows a high correlation.

As neutron log reads a lower value, the PEF value
increases and the rock tends to be more condensate,
with PEF reading 5 belongs to anhydrate and limestone.
On the other hand, as neutron log increases, the rocks is
more porous and the PEF value decreases which is
between 2 and 3 for porous sandstones and shale. The
negative coefficient of the regression line indicates an
inverse relationship between neutron and PEF logs.
In Figure 10, the Cross-plot of PEF and density logs in
well A is shown. Density log data are plotted from 2 to
2.9 g/cm
2
in the horizontal axis and the colored data
refer to gamma ray from 0 to 200 API in which the
green and yellow points are clean sands, shaly sands
and shales respectively.
The linear relationship between RHOB and PEF logs is
PEF=-10.4896+5.59337RHOB in which CC= 0.78
shows a high correlation. AS the density log increases,
the rocks become more condensate and the PEF value
increases too, which refers to limestone and anhydrate.
As the density log decreases, the porosity of formation
increases.
The PEF value decreases to 2 and 3 for porous
sandstones and shale. The positive coefficient of the
regression line indicates a direct relationship between
neutron and PEF logs.
Figure 11 is the Cross-plot of PEF and NDS logs in well
A. The horizontal axis is the NDS log from -10 to 10
and the PEF log is plotted on the vertical axis from 0 to
10 and the colored data are gamma ray log. The NDS
log value is negative in sandstones.

Figure 8. PEF and DT logs cross plot in well A Figure 9. PEF and NPHI logs Cross plot in well
Figure 10. PEF and RHOB logs Cross plot in well A Figure 11. PEF and NDS logs Cross plot in well A



980
The NDS log value is positive in Anhydrite, dolomite
and shale and near 0 where pure lime and the PEF log
reads 2 in sandstones. The linear relationship is as PEF=
3.681140+3046690 * NDS and CC=0.50, which shows
a good correlation.
The NDS log is the separation of the neutron density.
In order to be separated, they should be normalized and
shown with the same scale. The equation below is used
for this purpose (Liu, 2012):
NDS=(RHOB-1.95)/0.06-(0.45-NPHI)/0.03
(3)
Figure 12, the Cross-plot of the estimated PEF log and
the sonic log in well D is drawn with sonic log data
plotted in the horizontal axis from 40 to 140
microseconds. The PEF log data are drawn from 0 to 10
in the vertical axis. Points where the transit time
decreases show a more condensate rock such as
limestone and anhydrate.
At points where the transit time increases, the formation
is more porous which tends to be sandstone and shale
with green to yellow colors. The estimated PEF log in
condensate formations reads 5 for porous rocks.

Additionally, where Sonic log transit time increases, the
estimated PEF log reads between 2 and 3 which is
related to sandstone and shale. This is an indicative of
the correctness of the estimated PEF log by the MRGC
method. The linear relationship between sonic log and
the PEF log is as: PEF =9.70276-0.0906589 * DT and
CC =0.82, which is a high correlation between DT and
PEF logs.
Figure 13 is the Cross-plot of the estimated PEF and the
neutron logs in well D in which the neutron log is
plotted in horizontal axis from -0.1 to 0.4 mV and the
estimated PEF log from 0 to 10. As the neutron log
increases, the porosity increases as well and the
estimated PEF log reads a low value between 2 and 3
which is related to sandstone and shale. As the neutron
log decreases, the PEF value reads between 5 and 6.
The range between 5 and 6 shows a dense and less
porous formation which is related to limestone and
Anhydrite. This indicates that the estimated PEF log by
the MRGC method is acceptable. The linear relationship
between NPHI and PEF logs is as: PEF= 5.80432-
14.7678 * NPHI in which CC=0.69 .

Figure 14 the Cross-plot of the estimated PEF and the
density logs in well D is drawn with the density log data
plotted in the horizontal axis from 2 to 2.9 g/cm
2
and the
PEF log data are drawn from 0 to 10 in the vertical axis.
As the density is increased, the formation becomes less
porous and more condense and the estimated PEF log
reads
between 5 and 5.5 which relates to limestone and
anhydrate.
As the Density log reading decreases, the PEF reading
is between 2 to 3 which shows porous formations like
sandstone and shale rocks .This shows that the
estimated PEF log by the MRGC method is excellent.
The linear relationship between RHOB and PEF logs is
Figure 13. PEF and NPHI logs Cross plot in well D Figure 12: PEF and NPHI logs Cross plot in D well



981
PEF=-11.53916.07501 with a high correlation of CC =
0.80.
Figure 15 is the Cross-plot of the estimated PEF and
NDS in well D. The NDS log is plotted in horizontal
axis from -10 to 10 and the estimated PEF log from 0 to
10. In NDS log, the negative points show sandstones
and the estimated PEF log reads 2. The linear
relationship between NDS and PEF logs is PEF
3.669210.235632 with CC=0.41.

CONCLUSION
To identify formation lithology, it is required to have
PEF and LITH logs. However, these logs are not
available in old wells. Additionally, there is a
limitation of core data and lack of coring in all the
wells for lithology identification purposes.
In this study, the MRGC method is performed to
estimate well logs. In this method, petrophysical data
are grouped into a number of clusters based on NI
and KRI indices. Compared to conventional methods,
the introduced procedure is robust and shows
superior results.
This type of clustering is the advantage of the MRGC
method in respect to other clustering techniques. In
each obtained cluster from the MRGC method, a
relationship between PEF log and the model logs and
also, between LITH log and the model logs is
derived.
In this study, the formation is divided into 15
different clusters using MRGC method, in which, for
each cluster the majority of PEF and LITH logs are
derived and the average value of each cluster is
related to PEF and LITH logs. For each average
value, a number is defined which is the estimated
PEF and LITH, so the MRGC method is more
accurate. As the number of clusters are increased, the
estimated PEF and LITH logs are smoother.
References
1. Bobick, A., Intille, S., Davis, J., Baird, F.,
Pinhanez, C., Campbell, L., Ivanov, Y., Schtte,
A., and Wilson, A., 1999, The KidsRoom: a
perceptually based interactive and immersive
story environment, RESENCE: Teleoperators
Virtual Environ. 8, 367391.
2. Contea, D., Foggiab, P., Jolionc, J .-M., and
Ventoa, M., 2006, AGRaph-based, multi-
resolution algorithm for tracking objects in
presence of oCClusions, Pattern Recognition 39,
562 572.
3. Liu Chengbing, 2012, SYSTEM AND METHOD
FOR SWEET ZONE IDENTIFICATION IN
SHALE GAS RESERVOIRS, Earth science well
logging or borehole study by induction or
resistivity logging tool, Patent application
number: 20120065887.
4. Olson, T.J ., and Brill, F.Z., 1997, Moving object
detection and event recognition algorithm for
smart cameras, Proceedings of DARPA Image
Understanding Workshop, 159175.
5. Sutadiwirya, Y., Abrar, B., Henardi, D.,
NuGRoho, B. H., and Wibowo, R. A., 2008,
Using MRGC (Multi Resolution GRaph-Based
Clustering) Method to InteGRate Log Data
Analysis and Core Facies to Define Electrofacies,
in the Benua Field, Central Sumatera Basin,
Figure 14. PEF and NPHI logs Cross plot in well D
Figure 15. PEF and NDS logs Cross plot in well D


982
Indonesia, International Gas Union Research
Conference, IGRC, Paris.
6. Ye, S.J., and Rabiller, Ph., 2000, A New Tool For
ElectroFacies Analysis: Multi-Resolution GRaph-
Based Clustering, paper, SPWLA 41st Annual
Logging Symposium, June 4-7.

10/20/2012

FacimagePaperEx-150 10688life0904 974 982

Enviado por

Dados do documento

Descrição original:

Título original

Direitos autorais

Formatos disponíveis

Compartilhar este documento

Compartilhar ou incorporar documento

Opções de compartilhamento

Você considera este documento útil?

Este conteúdo é inapropriado?

Direitos autorais:

Formatos disponíveis

FacimagePaperEx-150 10688life0904 974 982

Enviado por

Direitos autorais:

Formatos disponíveis

Life Science J ournal 2012;9(4) http://www.lifesciencesite.

Você também pode gostar